An 8-gene signature for prediction of prognosis and chemoresponse in non-small cell lung cancer

Abstract

Identification of a potential gene signature for improved diagnosis in non-small cell lung cancer (NSCLC) patient is necessary. Here, we aim to establish and validate the prognostic efficacy of a gene set that can predict prognosis and benefits of adjuvant chemotherapy (ACT) in NSCLC patients from various ethnicities. An 8-gene signature was calculated from the gene expression of 181 patients using univariate Cox proportional hazard regression analysis. The prognostic value of the signature was robustly validated in 1,477 patients from five microarray independent data sets and one RNA-seq data set. The 8-gene signature was identified as an independent predictor of patient survival in the presence of clinical parameters in univariate and multivariate analyses [hazard ratio (HR): 2.84, 95% confidence interval CI (1.74-4.65), p=3.06e-05, [HR] 2.62, 95% CI (1.51-4.53), p=0.001], respectively. Subset analysis demonstrated that the 8-gene signature could identify high-risk patients in stage II-III with improved survival from ACT [(HR) 1.47, 95% CI (1.01-2.14), p=0.044]. The 8-gene signature also stratified risk groups in EGFR-mutated and wild-type patients. In conclusion, the 8-gene signature is a strong and independent predictor that can significantly stratify patients into low- and high-risk groups. Our gene signature also has the potential to predict patients in stage II-III that are likely to benefit from ACT.

Keywords: chemosensitivity; microarray analysis; non-small cell lung cancer; prognosis.

Figure 1. Survival analysis of the training data set

A. Schematic overview of the procedure used to construct the 8-gene signature based on gene expression data. B. The relative prognostic index based on the 8-gene signature expression of each patient. C. The heatmap of the median centered 8 genes’ expression profiles (red, relative high expression; green, relative low expression) between low- and high-risk groups. D. Kaplan-Meier plots for OS of two risk groups in the training data set. The p values were computed by log-rank test.

Figure 2. Validation of the 8-gene signature classifying independent data sets

A. Schematic overview of the strategy used for the construction of the prediction model and evaluation of predicted outcomes in five independent data sets by the 8-gene signature. B. All combined validation data sets. C-G. GSE31210, GSE30219, GSE29013/E-MTAB-923, GSE68465, and GSE42127 were classified by the 8-gene signature into low- and high-risk groups, and evaluated by Kaplan-Meier analyses. The p values were computed by log-rank test.

Figure 3. Kaplan-Meier survival analysis of the 8-gene signature with stages

A. Patients in all stages in the combined training and validation data sets. B. Patients in stage I in the combined training and validation data sets. C. Patients in stage II in the combined training and validation data sets. D. Patients in stage III in the validation data sets were classified by the 8-gene signature into low- and high-risk groups. The p values were computed by log-rank test.

Figure 4. Kaplan-Meier survival analysis of the 8-gene signature with adjuvant chemotherapy

Patients from combined validation data sets with available ACT data were included for analysis. A-B. Patients in high-and low-risk groups with chemotherapy in stage II-III. C-D. Patients in high- and low-risk groups with chemotherapy in stage III. Patients were plotted according to presence and absence of ACT. The p values were computed by log-rank test.

Figure 5. Kaplan-Meier survival analysis of the 8-gene signature with gene mutations

A. Kaplan-Meier curves of patients in EGFR. B. KRAS in the validation data sets. Each group was classified by the 8-gene signature into low- and high-risk groups. The p values were computed by log-rank test.

Figure 6. Kaplan-Meier survival analysis of the 8-gene signature with histological subtype

A. Kaplan-Meier curves of patients with adenocarcinoma (ADC). B. Kaplan-Meier curves of patients with squamous cell carcinoma (SQC). Each group was classified by the 8-gene signature into low- and high-risk groups. The p values were computed by log-rank test.